Developing a hybrid system for stock selection and portfolio optimization with many-objective optimization based on deep learning and improved NSGA-III

Information Sciences, Journal Year: 2024, Volume and Issue: 670, P. 120549 - 120549

Published: April 4, 2024

Language: Английский

---

FE-RNN: A fuzzy embedded recurrent neural network for improving interpretability of underlying neural network

Information Sciences, Journal Year: 2024, Volume and Issue: 663, P. 120276 - 120276

Published: Feb. 9, 2024

Deep learning enables effective predictions. But deep structures face some challenges on human interpretability compared to conventional techniques, e.g., fuzzy inference systems. It motivates more research works alleviate the black box nature of with performance maintained. This paper proposes a fuzzy-embedded recurrent neural network (FE-RNN) improve underlying networks. is parallel structure comprising an RNN and Pseudo Outer-Product based Fuzzy Neural Network (POPFNN) that share common set input output linguistic concepts. The processes undertaken are associated by using rules in embedded POPFNN. IF-THEN provide better process hybrid allows realisation data driven implication modelling entailment within networks (FNN) structure. FE-RNN obtains consistent results than other FNN experiment Mackey-Glass dataset. achieves about 99% correlation for forecasting prices market indexes. Its also discussed. then acts as prediction tool financial trading system forecast-assisted technical indicators optimised Genetic Algorithms. outperforms benchmark strategies experiments.

Language: Английский

---

An enhanced Transformer framework with incremental learning for online stock price prediction

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(1), P. e0316955 - e0316955

Published: Jan. 13, 2025

To address the limitations of existing stock price prediction models in handling real-time data streams—such as poor scalability, declining predictive performance due to dynamic changes distribution, and difficulties accurately forecasting non-stationary prices—this paper proposes an incremental learning-based enhanced Transformer framework (IL-ETransformer) for online prediction. This method leverages a multi-head self-attention mechanism deeply explore complex temporal dependencies between prices feature factors. Additionally, continual normalization is employed stabilize stream, enhancing model’s adaptability changes. ensure that model retains prior knowledge while integrating new information, time series elastic weight consolidation (TSEWC) algorithm introduced enable efficient training with incoming data. Experiments conducted on five publicly available datasets demonstrate proposed not only effectively captures information but also fully exploits correlations among multi-dimensional features, significantly improving accuracy. Notably, shows robust coping frequently changing financial market

Language: Английский

---

Predicting Economic Trends and Stock Market Prices with Deep Learning and Advanced Machine Learning Techniques

Electronics, Journal Year: 2024, Volume and Issue: 13(17), P. 3396 - 3396

Published: Aug. 26, 2024

The volatile and non-linear nature of stock market data, particularly in the post-pandemic era, poses significant challenges for accurate financial forecasting. To address these challenges, this research develops advanced deep learning machine algorithms to predict trends, quantify risks, forecast prices, focusing on technology sector. Our study seeks answer following question: “Which supervised are most efficient predicting economic trends under what conditions do they perform best?” We focus two recurrent neural network (RNN) models, long short-term memory (LSTM) Gated Recurrent Unit (GRU), evaluate their efficiency industry prices. Additionally, we integrate statistical methods such as autoregressive integrated moving average (ARIMA) Facebook Prophet like Extreme Gradient Boosting (XGBoost) enhance robustness our predictions. Unlike classical algorithms, LSTM GRU models can identify retain important data sequences, enabling more experimental results show that model outperforms terms prediction accuracy training time across multiple metrics RMSE MAE. This offers crucial insights into predictive capabilities techniques forecasting, highlighting potential XGBoost price

Language: Английский

---

Accurate Stock Price Forecasting Based on Deep Learning and Hierarchical Frequency Decomposition

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 49878 - 49894

Published: Jan. 1, 2024

The stock market plays an increasingly important role in the global economy. Accurate price forecasting not only aids government predicting economic trends, but also helps investors anticipate higher expected returns. Nevertheless, hurdles such as non-linearity, complexity and high volatility make it a daunting task to predict prices. To address this issue, paper proposes new hybrid model, termed Hierarchical Decomposition based Forecasting Model (HDFM), decompose forecast prices hierarchical fashion. model utilises complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) for initial of time series. enhance predictive efficiency, sub-series similar sample entropy from are combined K-means clustering method. Through thorough analysis, is found that first contains more high-frequency signals. Therefore, subjected second variational (VMD). Afterwards, gated recurrent unit (GRU) used each individually. final results obtained by merging prediction outcomes. proposed has been evaluated on three different markets. experimental showed outperformed other methods across all indices. Moreover, ablation studies demonstrated effectiveness individual component within model.

Language: Английский

---

A Novel Integrated Approach for Stock Prediction Based on Modal Decomposition Technology and Machine Learning

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 95209 - 95222

Published: Jan. 1, 2024

After the COVID-19 ended, global economy gradually recovered. Due to nonlinearity, complexity, and high noise of financial time series, stock price prediction has become one most challenging tasks in market. To tackle this challenge enhance performance complicated markets, we propose a novel integrated approach based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Long Short-Term Memory (LSTM), ensemble learning algorithm LightGBM simultaneously improve fitting accuracy prediction. In addition, prevent overfitting predictive performance, study adopted Simulated Annealing (SA) for optimization. The proposed hybrid model is comprehensively evaluated by comparing it single LSTM, RNN, other popular models. Three evaluation metrics, namely Root Mean Square Error (RMSE), Absolute (MAE), accuracy, are used compare aforementioned experimental results indicate that CEEMDAN-LSTM-SA-LightGBM outperforms all comparative models better accuracy.

Language: Английский

---

Time series forecasting of stock market indices based on DLWR-LSTM model

Finance research letters, Journal Year: 2024, Volume and Issue: 68, P. 105821 - 105821

Published: July 11, 2024

Language: Английский

---

Machine Learning-Based Process Optimization in Biopolymer Manufacturing: A Review

Polymers, Journal Year: 2024, Volume and Issue: 16(23), P. 3368 - 3368

Published: Nov. 29, 2024

The integration of machine learning (ML) into material manufacturing has driven advancements in optimizing biopolymer production processes. ML techniques, applied across various stages production, enable the analysis complex data generated throughout identifying patterns and insights not easily observed through traditional methods. As sustainable alternatives to petrochemical-based plastics, biopolymers present unique challenges due their reliance on variable bio-based feedstocks processing conditions. This review systematically summarizes current applications techniques aiming provide a comprehensive reference for future research while highlighting potential enhance efficiency, reduce costs, improve product quality. also shows role algorithms, including supervised, unsupervised, deep

Language: Английский

---

Diagnosis of Alzheimer’s disease using FusionNet with improved secretary bird optimization algorithm for optimal MK-SVM based on imaging genetic data

Cerebral Cortex, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 4, 2025

Alzheimer's disease is an irreversible central neurodegenerative disease, and early diagnosis of beneficial for its prevention intervention treatment. In this study, we propose a novel framework, FusionNet-ISBOA-MK-SVM, which integrates fusion network (FusionNet) improved secretary bird optimization algorithm to optimize multikernel support vector machine diagnosis. The model leverages multimodality data, including functional magnetic resonance imaging genetic information (single-nucleotide polymorphisms). Specifically, FusionNet employs U-shaped hierarchical graph convolutional networks sparse attention select feature effectively. Extensive validation using the Disease Neuroimaging Initiative dataset demonstrates model's superior interpretability classification performance. Compared other state-of-the-art learning methods, FusionNet-ISBOA-MK-SVM achieves accuracies 98.6%, 95.7%, 93.0%, 91.8%, 93.1%, 95.4% HC vs. AD, EMCI LMCI EMCI, LMCI, respectively. Moreover, proposed identifies affected brain regions pathogenic genes, offering deeper insights into mechanisms progression disease. These findings provide valuable scientific evidence preventive strategies

Language: Английский

---

Damage Detection for Truss Bridge Structure Using XGBoost

Advances in science and technology, Journal Year: 2025, Volume and Issue: 158, P. 65 - 74

Published: Jan. 6, 2025

Structural health monitoring (SHM) is a burgeoning area of interest among modern research endeavors, motivated by the application state-of-the-art machine learning models. During last few years, many researchers have proposed techniques for analysis SHM datasets, particularly those corresponding to sequence data collected from sensors. Following flow this research, in work, we introduce an effective approach utilizing eXtreme Gradient Boosting (XGBoost), potent ensemble framework rooted gradient boosting damage detection. A dataset cases Nam O bridge, steel truss bridge railways, applied assess damages. To evaluate effectiveness method used, common DL models such as One-Dimensional Convolutional Neural Network (1DCNN) and Long Short-Term Memory (LSTM) are also considered. Moreover, influence round on overall result will be analyzed. The results validation set test both illustrate that XGBoost performs better accuracy than 1DCNN LSTM with 100% 95.7%, respectively. Besides, model achieved lowest mean square error (MSE) only 4.3% set. These demonstrate significant potential structures, especially through utilization time-series data.

Language: Английский

---